When considering data center networking architecture (the layout of the cabling infrastructure and the way servers connect to switches) designers must strike a balance between reliability, performance, agility, scalability and cost. And, to optimize data center investment, the architecture must also be able to support both current and future speeds and applications.
So what should designers consider? In our post we detail three key decision factors when selecting data center architecture:
Centralized – This is best for smaller data centers that are under 5,000 square feet. The data center hosts separate LAN/SAN environments which provide very efficient utilization of port switches and makes it easy to add and manage components. But it is important to note that this layout does not scale up well, making it difficult to support expansions.
Zoned – This layout is ANSI/TIA-942 Data Center Standards recommended, as it is very scalable, repeatable and predictable. A zoned layout consists of distributed switching resources, so switches can be distributed in an end-of-row (EoR) or middle-of-row (MoR) location. In some scenarios, EoR switching can provide performance advantages when the LAN ports of two servers that exchange large volumes of information are placed on the same EoR switch to take advantage of the low latency of port-to-port switching.
A potential disadvantage of EoR switching would be the requirement to run cable back to the EoR switch. However, assuming every server is connected to redundant switches, this cabling can exceed what is required in top-of-rack architecture.
Top of Rack (ToR) – This type of switching layout typically consists of two or more switches in each server cabinet. This layout can be a good choice for dense one-rack unit server environments; in this design, switches are placed at the top of the rack and all servers are cabled to both switches for redundancy. Yet although cabling is utilized more efficiently in top-of-rack designs, this layout can be more expensive due to the cost of switches and under-utilization of ports.
Cross-connect – This connection method connects cabling runs, subsystems and equipment using patch cords and jumpers that attach to connecting hardware at each end. The cross-connect method is easily scaled up and flexible due to an “any-to-any” concept. This works well as there is no need to disturb the electronic ports or backbone cabling when making the connection, although this concept can be more expensive to install as it requires more cabling.
Interconnect – Interconnect brings a patch cord directly from the electronics port to the backbone cabling. This solution requires fewer components making this approach more cost-effective. However it does reduce flexibility and add risk, as users must directly access the electronic port to make the connection. Consequently, CommScope recommends using the cross-connect method for greater solution flexibility and efficiency in the data center.
In planning the data center’s LAN/SAN environment (and taking into account new technologies such as 25/40/100GbE, 40GFcoE and Fiber Channels), designers must understand the benefits and limitations of each data center architecture application, for best performance and future-proofing. They must then select an architecture that will not only support current applications but also have the ability to migrate to higher-speed future applications.
You can learn more about data center technologies, challenges, architectures and solutions here.